How to translate contract images? How to preserve the layout of scanned documents

Key points of this article

PDFs that can select text can be translated directly. Scans or photos that cannot be selected must be OCRed first, otherwise the translation will have missing paragraphs or typos.
After OCR recognition, the amount, date, number and name must be checked against the original. Errors in these contents will directly affect the meaning of the contract.
Select bilingual comparison to facilitate paragraph-by-paragraph verification, and select translation only to facilitate printing and archiving. You can produce one copy of each.
Legal contracts should use legal-style translation settings and be reviewed by professionals before official use.

The contract sometimes consists of just a photo, or a PDF that has been scanned so crookedly that the text cannot be reproduced. If you hand it over directly to the translation tool, the common result is that the paragraph cannot be recognized, the table is misplaced, or even the entire page is missing from translation. This article is written for people who need to deal with this type of contract, such as those who receive a photo version of the agreement from an overseas partner and need to translate the scanned copy into another language. After reading, you will be able to judge which path your document should take, complete the step-by-step process of "first identify, then translate, and retain format", and know what content to check before delivery.

The key to this kind of file is the order: first turn the text in the picture into selectable text, and then do the translation. If you skip recognition and translate directly, the translation is usually incomplete and the format is easily messed up. The judgment methods and practices are first given below.

The contract only has pictures or scanned copies. How can it be translated so that the format can be retained? One sentence conclusion

First confirm whether the text can be selected in the file: if it can be selected, translate it directly; if it cannot be selected, do OCR recognition first and then translate it. When translating, choose a style that is consistent with the purpose of the contract, and keep the bilingual comparison or translation format for easy verification.

your situation	Recommended practices	reason
One or several photos taken with your mobile phone	First organize it into a PDF, then do OCR, and finally translate	The photo has perspective distortion and shadows. Convert it to PDF first and correct it to make the recognition rate more stable.
Scanned PDF, Chinese characters cannot be selected	OCR the PDF on the desktop client and upload the translation after checking	First get the PDF with optional Chinese text, and the paragraph and table structure will be more complete.
The scanned text is clear and you only need to understand the general idea.	Upload directly to omnidoc.top translation, select bilingual comparison	When formal use is not needed, one less identification step can be saved
The contract must be signed, reported or used as evidence	After OCR, check item by item and submit it to a professional translator or lawyer for review.	Both machine translation and recognition may be subject to error, and the original must prevail for official purposes.
There are many files in different formats	Separately process documents by type and check each document individually	It is easy to miss a certain part of the recognition error when mixed together.

Before you start: Confirm these points first

Judging a few things before taking action can save a lot of rework time.

First confirm whether the document is a photo or a scan. Try dragging a section of text with the mouse in the PDF reader. If it can be selected, it means that the file itself has a text layer and can be translated directly; if it cannot be selected, it means that it is a full-page image and needs to be recognized first. Photo files, usually JPG or PNG, also need to be organized into PDF first.

Second, look at the clarity and shooting angle. OCR can easily misrecognize numbers and punctuation on pages with blurred, creased, slanted or shadowed writing. When taking photos, try to face the page directly and the lighting is even; if the original is very blurry, it is recommended to find a clearer version.

Third, clarify the purpose of the translation. Just understand the content, bilingual comparison is convenient; to print, file or submit to the other party, you need a neat translation-only version. Different uses require different output methods.

Fourth, clarify the type of contract. Legal texts such as labor contracts, lease contracts, confidentiality agreements, etc. are very particular about the wording and sentence structure. It will be more suitable to choose the legal style translation setting. Especially those involving amount, liability, and jurisdiction clauses must be specifically checked.

Fifth, pay attention to numbers and proper nouns in the contract. Amount, date, ID number, company registration number, name and address are parts of OCR that are prone to errors and directly affect the validity of the contract. Check them first when checking.

Steps to translate a scanned contract in OmniDoc

The following uses a 12-page scanned contract as an example to illustrate the complete process.

Arrange the original documents. If there are multiple photos, sort them by page number first and merge them into one PDF file. It is recommended to include the page number in the file name so that you can check whether the page number is complete later.
Open the PDF on the desktop client and do OCR. Use the OCR function in the PDF editor to recognize the text on the scanned page. Once completed, the text should be selectable and searchable.
Check the recognition results page by page. Compare with the original, focusing on the amount, date, serial number, name and clause number. Find errors and modify paragraph text directly in the PDF editor.
Export or save the recognized PDF. Confirm that each page has Chinese text selectable, and then use this file as the source file for translation.
Open omnidoc.top and upload the file. Upload the recognized PDF and the system will process the entire file.
Select translation settings. Select legal style for contract documents. The output method can be selected as bilingual comparison to facilitate paragraph-by-paragraph comparison; if a finished product is required, a separate translation-only version can be generated.
Download the results and open them for verification. Check whether the translation retains the page structure, paragraph order and table placement of the original text.
If used in formal occasions, please submit it to a professional for review. Mark the terms that need to be confirmed on the translation, and send it to the translator or lawyer for final confirmation before use.

Check each item after completion

After you get the translation, don't send it directly. Check item by item according to the list below.

Check item	What do you think?	What if there's a problem?
Whether the number of pages is complete	Compare the page number of the original to see if the last page of the translation corresponds to the last page of the original.	Go back to the source file when the page is missing and check to see if some pages have been missed when taking photos or merging.
Amount and date	Compare the number, uppercase and lowercase amount and date format in the original one by one	If you find an error, correct the recognized text in the source file and retranslate it.
Serial number and clause serial number	See if the serial number of the clause (such as Article 1, Article 2) and the attachment number are continuous and consistent with the original.	Misalignment of serial numbers is usually caused by missing lines in recognition, which can be completed in the OCR result.
Layout and tables	Check whether the rows and rows of the form are aligned, and whether the signature column and the header and footer are in a reasonable position.	When the layout is out of order, first confirm whether the source file is the PDF of the optional text, and then reprocess it.
Proper noun	Check whether the translation of the company's name, person's name and address is consistent	Manually modify the unification and retain the original name in the official document
Signature bar and seal area	To see if the position of the signature and seal is blocked or shifted.	The signature of the original should not depend on the translation, which indicates "the original shall prevail"

# which method is suitable for you?

If you just want to understand the general idea of the contract, it will be enough to upload and translate the scanned copy directly and quickly browse it in bilingual control. You can skip the manual correction of OCR at this point, but pay attention to the key numbers.

If you need to send the translation to the other party, or if you want to keep the file, you should go through the complete process: first OCR, item by item, then generate the translation-only version, and finally have the professional review it. Once the contract involves amount and liability, even a decimal point error may lead to disputes.

If there are a large number of forms, attachments and handwritten notes in the contract, spend more time checking the layout after OCR. The recognition effect of handwritten content is usually poor, so it is recommended that the handwritten part be manually confirmed separately.

# frequently asked questions (FAQ)

1. Can the contract photos taken by mobile phone be translated directly?

You can upload images for translation first, but photos usually have perspective distortion and shadows. It is recommended to organize them into PDF and do OCR on the desktop client. Check the recognition results and then translate them. The results will be more reliable.

2. Will the form be misplaced after the translation of the scanned piece?

Most of the table dislocation comes from the recognition stage. The layout is more complete when the source file is the PDF of optional text. After translation, you should check the rows and rows of the tables page by page, and if you find the dislocation, you will go back to the source file for processing.

3. Can the translation result be directly used as an official translation?

It's not recommended. Both machine translation and OCR can make mistakes, especially the legal terms and amounts. The official use shall be based on the original and shall be reviewed by a professional interpreter or lawyer.

4. Can you also recognize the handwritten content in the contract?

The recognition effect of handwritten content is usually not as stable as that of print. If you encounter handwritten numbers, signatures or annotations, it is recommended that you confirm them manually against the original, rather than relying solely on the recognition results.

5. Which should be chosen, bilingual control or translation only?

When checking, choose bilingual control, paragraph by paragraph is more convenient; when you need to print, archive or send to each other, then generate a translation-only version. The two copies can be kept separately to avoid mixing.

# Summary

The key to the contract translation of the picture is to identify the text in the picture, then translate it, and check the number, number and layout before delivery. According to the order of "judging the file type, OCR identification, item-by-item inspection, and choosing the appropriate output mode", scanning the contract can also maintain a clear page structure. Before formal use, please take the original and submit it to a professional for review.

Open OmniDoc to identify and translate scanned parts